A market can finish its work while leaving the world intolerable. Every bargain worth making has been made; every improvement that could help someone without hurting anyone has been exhausted; and still one household has plenty and another scarcely enough. The contradiction is only apparent. The theorem that certifies the market's efficiency was never a theorem about the decency of its result. It tells us that a particular kind of waste has ended. It does not tell us that the people standing at the end have received what they ought.
This is the distinction the first welfare theorem makes exact, and the distinction most easily lost when the result is described as the largest possible pie. In the usual private-goods setting, consumers choose what they prefer from the bundles their budgets permit; firms choose production to maximize profit at prices they take as given; markets clear; and relevant costs and benefits are represented in the trading system. With the appropriate condition that consumers can always find a preferred bundle arbitrarily nearby, a competitive equilibrium is Pareto efficient. No feasible rearrangement can make at least one person better off while leaving everyone else at least as well off. The definition compares each person with that same person in another arrangement. It does not add their pleasures together, compare their hunger, or measure the deservingness of their claims.
The strength of the result lies in the way a common set of prices brings separate decisions together. A consumer who has chosen the best affordable bundle cannot be offered something strictly better at a cost she can already meet. A firm maximizing its profit cannot supply a more valuable net output from its feasible alternatives. If a proposed rearrangement made everybody at least as well off and somebody better, its combined demands would, under the theorem's conditions, require more purchasing power than the resources and production plans could furnish at those same prices. What seemed an opportunity for the whole would contradict the choices that support the equilibrium. That is more than a story about a kindly merchant. It is a consistency result linking millions of budgets and production decisions without appointing one person to rank everybody's satisfactions.
It is also less than the proposition that the economy has maximized one grand total of money gain. There is a setting in which that familiar picture is exact: let buyers' benefits and sellers' real costs be expressible in a common money unit, let income effects be absent over the choices being compared, and count every affected party. Then the sum of benefits less costs is the surplus, and a transfer between buyer and seller changes its division without changing its amount. A unit worth more than it costs belongs in the total; one costing more than it is worth does not. The diagram of the largest area between demand and supply is powerful because it makes this special case visible. The general welfare theorem is framed more carefully. It rules out an improvement for someone at nobody else's expense; it does not supply a common ruler with which to rank all the different efficient worlds.
Imagine, to make the silence unmistakable, an invented economy containing two people and six divisible loaves. There is no production, no spoilage, no other good, and each person strictly prefers more bread. Five loaves for one and one for the other is efficient: giving the second person more must take bread from the first. Three apiece is efficient for exactly the same reason. So is any other allocation that uses all six. At a positive bread price, if each starts with the allotment in question and can spend only its value, each can afford precisely that allotment; the unequal and the equal arrangements can both be supported as competitive equilibria. Nothing in the example establishes which division is just. Its point is that the certificate of efficiency cannot distinguish them. The theorem can approve the absence of waste in an arrangement whose distribution we have independent reasons to reject.
The second welfare theorem appears to offer a bridge between those reasons and the market's powers. Under stronger conditions, including suitable convexity, an efficient allocation selected on distributive grounds can be supported by competitive prices after an appropriate redistribution of purchasing power. The result gives distribution a place before the trading begins. Decide which efficient outcome is wanted, supply the wealth holdings that support it, and allow decentralized choices to do their work. The market need not be asked to invent the social objective; it may implement an objective chosen elsewhere. This is an important possibility. It also makes clear how much the phrase leave the market alone conceals: the initial claims are part of the machinery, and changing them can change which efficient outcome the machinery delivers.
There is work hidden inside that redistribution. A transfer fixed independently of what its recipient subsequently earns or buys need not change the relative prices at which choices are made. A tax that rises with earnings, by contrast, can change the reward to another hour's work; a subsidy conditioned on a purchase can change the purchase. It does not follow that every actual redistribution must be wasteful, still less that existing inequality is efficient. It follows that an institution implementing a transfer has effects of its own, which must be examined with the transfer. Information, administration, incentives, and the treatment of people at the boundary are part of the proposed arrangement. A theorem that permits a clean separation under specified conditions cannot guarantee that a particular government's instruments will achieve it.
The same discipline applies to the first theorem's failures. Smoke falling on a neighbor without a price, a seller able to hold back output to raise its own return, or an important risk for which no suitable contract can be made changes the problem. The price system no longer represents the full set of consequences in the way the proof requires. That gives a reason to investigate the particular failure and the remedies available. It gives no automatic certificate to whichever remedy is proposed. Nor does the theorem say how swiftly a market reaches equilibrium, whether it reaches it at all, or whether a bad outcome observed today is explained by the failure of one assumption rather than another. A benchmark earns its force by identifying the missing comparison, not by furnishing a verdict before the comparison has been made.
The original promise therefore survives, but with its object restored. There is a real achievement in arranging production and exchange so that a benefit is not abandoned when it could be obtained without anyone's sacrifice. There is a real loss when power blocks such an improvement merely to enlarge the power-holder's share. Yet the refusal of waste and the refusal of injustice are different reasons for action, and they can meet in the same policy without becoming the same reason. An argument for a different distribution need not show that the present one wastes bread. It can say, plainly, that the bread ought to be differently shared.
The theorem tells us when no one can gain without someone losing. It does not tell us when a loss is justified, or whose claim should prevail.
Does the largest pie excuse any slicing? Even the question gives the theorem a larger jurisdiction than it possesses. In its general form it does not rank every pie on a single scale; in the simpler setting where a money total can be maximized, that total still does not allocate the entitlement to it. The competitive result can be an extraordinary achievement and an inadequate settlement at once. The next plate follows the party who understands that distinction perfectly: the holder of power, interested less in the certificate that the surplus exists than in the prize that can be kept.